How Future Value May Lie In Making Driverless Cars Cooperate

Power and profits will accrue to those who design - and own - the interface with other technologies that enable the distribution of content and effectuate sales. JL

Kim Tingley reports in the New York Times:

An intelligent automated future will allow vehicles to drive themselves. The other technology will connect vehicles to their surroundings and to one another, in order to optimize the functioning of the vehicular ecosystem as a whole. Those who play the largest roles in solving (that) will be in a position to dictate  which people and places they will empower and which they will upend.

In his 31-year career delivering frozen patties, ketchup packets, ice cream and such to McDonald’s, Baskin-Robbins and other restaurants across the country, Dave Mercer logged more than three million miles, many of them in rigs he termed “bad news”: boxy, twin-stick-shift trucks that belched smoke and had no heating, air conditioning or radio. But they were sleek compared with the 18-wheelers his father drove before he was born. His father used to say that going uphill he could walk faster than his load; sometimes he would stand out on the running board to cool off and steer one-handed through his open door. “You’re not going to stop technology,” Mercer told me, when I asked him why he had decided to finish his professional life by test-driving trucks full time for a four-year-old Silicon Valley start-up. We were standing on the side of I-280 in Mountain View, Calif.; Mercer, wearing a Hawaiian shirt and dark glasses, had just finished demonstrating a system that a company called Peloton Technology plans to introduce in commercial fleets next year. “I wanted to come over and be a part of this before I retired,” he said. “I can tell some of the younger drivers. ... ” In lieu of finishing his thought, he punched his co-pilot, a quiet computer engineer, in the shoulder. Hundreds of hours riding shotgun with Mercer — panoramic views of the horizon, cars flowing streamlike down below — had inspired him and several of the company’s other engineers to get a commercial driver’s license. “I want to be part of that, too,” Mercer said.

Someday cars will drive themselves and coordinate with one another and smart cities and roadways to eliminate traffic jams and accidents — including the fatal one that happens in the United States every 15 minutes — while reducing fuel consumption and effectively giving back to Americans an annual per-capita average of 335 hours in which, for better or worse, they can use their hands and eyeballs in other ways. When and in what shape this utopian future will arrive, though, remains a persistent mystery. What is certain is that those who play the largest roles in solving it will be in a position to dictate — as those who control infrastructure always have been — which people and places they will empower and which they will upend.

An intelligent automated future, however, depends on the maturation of two entirely separate sets of technologies. The first — and the one that has drawn most of the investment and media attention in the last five years — will allow vehicles to drive themselves without any input from their surroundings. The other set of technologies will connect vehicles to their surroundings and to one another, in order to optimize the functioning of the vehicular ecosystem as a whole. Ultimately, these two types of systems will have to complement each other, yet neither can be implemented all at once. In the meantime, the process of developing and deploying them often puts them at odds. Indeed, the two systems are, in a sense, philosophically opposed: libertarian self-reliance in a race with socialist collectivity to be the first to improve traffic flow.

In Silicon Valley and elsewhere, where numerous start-ups are competing with software giants like Google and Uber in the “self-driving” category, Peloton has left high-level automation out of its immediate business plan. Rather, it has staked its early success — and the money of investors like UPS, Volvo and Intel — on a cooperative technology known as “vehicle-to-vehicle communication,” or V2V, which makes use of a shortwave portion of the radio spectrum that the Federal Communications Commission set aside in 1999 primarily for traffic-safety messaging. A signal can zip through this bandwidth thousands of times faster than the second or two it might typically take a driver to react. This valuable swath of ether (worth billions to the telecommunications industry, which is currently lobbying the F.C.C. to “share” it) facilitates unlimited potential applications, not just for sending data among vehicles but also for syncing them with traffic lights and other features of the environment. Currently, vehicles using it — the vast majority belonging to research pilot projects — broadcast encrypted information 10 times per second to any receivers within 1,000 feet that can decode it. This means that, unlike vehicles that are automated but not connected, they can “see” beyond their actual field of view: an accident just over the crest of a hill, an oncoming car around the corner. “It will save lives, and everybody knows that,” says Harry Lightsey, the executive director for emerging-technologies policy at General Motors, maker of the 2017 Cadillac CTS, the first commercially available car in the United States to carry the technology. “A delay in rolling out V2V will cost lives, and that’s a tragedy.”

It has been nearly two decades since the F.C.C. allocated bandwidth to be used primarily by V2V. So why does talk about the future of transportation tend to be dominated by fully automated vehicles? One reason is that most people simply assume automation includes connectivity to others on the road. It is also hard to explain what V2V is: Like the internet, it describes a medium with many possible uses, not a stand-alone product. Nor is there a single definition in play, really; though V2V typically refers to systems that operate in the F.C.C.-dedicated bandwidth, it can technically mean any type of connection among vehicles, or between vehicles and infrastructure (V2I) or “everything” (V2X). Like the telephone’s, too, its value is contingent upon its dissemination — early customers, with no one to “call,” must buy into an abstraction.

Peloton believes that trucking companies are the natural place to start: They can outfit entire fleets at once and reap significant benefits from small improvements. Moreover, most of the 279 billion annual miles trucks travel are on highways, which are far more predictable than city streets and therefore an easier environment in which to usher in new driving features. Thus, the trucking industry is poised to be among the first to adopt both V2V and automated technologies on a grand scale, possibly becoming the arena in which they will compete with each other most ferociously for initial relevance and public acceptance.

On the day I met Mercer, he and Jake Gregory, a veteran F.B.I. agent who drove trucks in college and directed safety operations for a small fleet after leaving the bureau, were demonstrating Peloton’s first V2V application, called “platooning,” in a pair of 18-wheelers. (“Peloton,” the French word for “platoon,” also designates the main pack of cyclists in a road race.) They eased out onto I-280 with Mercer in the lead. I rode with Gregory, whose cab had been retrofitted with theater-style seats where a bunk bed used to be. These were occupied by Steve Boyd, a Peloton founder, who oversees external affairs; Oliver Bayley, the head of user experience design; and Rod McLane, the company’s marketing manager.

The specialties of these — user experience and public relations — were reminders of the “social problem” (as Bayley put it) that nags at Peloton. Their product is ready, and they expect to outfit hundreds of commercial trucks with it in 2018. But legislators, the general public and, crucially, truck drivers still fear it, often confusing it with automation, which many see as a threat to drivers’ jobs. (Even companies about to put it in their fleets decline to be named.) This unease is reflected in the rejection of the terms “self-driving” and “autonomous” in recent recommendations from SAE International, which publishes industry standards. Boyd, a former assistant press secretary for the White House, explained: “ ‘Autonomous’ is: You get into the truck and say, ‘I want to go to the office,’ and the truck says, ‘I want to go to the beach.’ ”

'It will save lives, and everybody knows that.'

Gregory flipped a switch on the dash, and a screen — mounted where a car’s rearview mirror usually is — blinked on. “Platooning not authorized in this area,” it said. The system, Gregory said, was calculating the weight of Mercer’s truck ahead: “It takes a few minutes for it to establish mass.” The software also used a cellular signal to check with a cloud-based server run by Peloton engineers to see if Gregory was in an appropriate location: a divided highway accessible by only on-off ramps, more easily negotiated than crossings.

The system chimed, and the message changed: “Ready to platoon.” Gregory pressed a button that allowed Mercer’s truck to take control of his throttle and braking systems over the V2V bandwidth. It began to reel Gregory’s truck in until they were exactly 66 feet apart. By maintaining that gap, it created a slipstream that reduced carbon emissions and saved fuel by an average of 7 percent: 10 percent for the follower and 5 percent for the leader. (A team led by researchers at Purdue University, with a grant from the United States Department of Energy, believes it can help Peloton increase the average to 20 percent, which the team estimates would be worth up to $20 billion a year to the trucking industry.)

While continuing to steer, Gregory removed both feet from the pedals and planted them on the floorboard. Mercer slowed at an exit ramp, and Gregory’s truck automatically did, too, beeping at him to resume control.

When I first heard about platooning, I imagined it would feel more death-defying, but though the size of the gap had initially unsettled Mercer and Gregory, it attracted little notice on the highway; oblivious passenger-car drivers frequently cut in, dissolving the platoon. In practice, the exercise felt more like an elaborate cruise-control upgrade than the opening bell in a transportation revolution. Bayley agreed: “It’s disappointing!”

Critics of V2V, especially those who are solely invested in full automation, dismiss platooning as mere high-speed tailgating. The major added safety benefits, they point out, come from the fact that, in order to platoon, trucks must be equipped with advanced safety features. Peloton readily agrees that its application, whose requirements include air disc brakes and state-of-the-art collision-avoidance, is an incentive for fleets to upgrade. V2V proponents, on the other hand, hope to leverage renewed federal interest in their systems to expand projects that have been gestating for decades. The Department of Transportation, under the outgoing Obama administration, proposed requiring that all new passenger vehicles be equipped with V2V — at least symbolically moving toward the realization of those goals. Earlier this month, however, reports suggested that the Trump administration might be inclined to shelve the proposal. “It’s important to get our government leaders to commit to this,” says Larry Head, an engineering professor at the University of Arizona, who is collaborating on V2I applications with Peloton. “We can’t wait five years. We’re ready now.”

To connect in any way is to cede control and reconfigure vulnerability. The history of civilization is one of trading pure independence — living alone in a cave, say — for comfort of one kind or another: companionship, enhanced productivity, protection. On balance, it is hard to argue that this exchange has not worked out in our favor. But the benefits of communing also exact steep prices — war, climate change, economic inequality — that are suffered individually. Since the completion of the first transcontinental railroad in 1869 transformed the American landscape and economy, public transportation and infrastructure have given physical shape to moral questions that arise when we are forced to choose between self-interest and the greater good. In 1967, the British philosopher Philippa Foot posed a thought experiment that became known as the “trolley problem.” A runaway trolley is headed for five people tied to its track. You are standing next to a lever that can divert the trolley to another track, where it will kill a single person. Is it more ethical to pull the lever or not?

As a society, we routinely make comparable decisions: when we let football players give one another brain damage in the interests of public entertainment; when a surgeon refrains from killing one terminal patient and transplanting his organs into five patients whom they might save. The trolley problem heightens individual culpability by bestowing absolute decision-making power on one person and demanding that she select not between actions but consequences, which in real life are rarely inevitable. Arguably, in doing so, the dilemma tends to reveal our natural inclination to accept fate over premeditation in situations with no happy outcomes, even if planning might partly mitigate an otherwise inevitable tragedy.

In recent years, the trolley problem has frequently been adapted to discussions about entirely automated cars, which respond to contingencies with binary logic. Suppose a distracted mother pushes a stroller with her baby in it out in front of such a vehicle — should the car be programmed to hit them or swerve into a tree? Azim Shariff, the director of the Culture and Morality Lab at the University of California, Irvine, was an author of a 2016 report in Science that described a series of surveys he and colleagues conducted in which participants overwhelmingly approved of autonomous vehicles, as the study terms them, that would “sacrifice their passengers for the greater good” and “would like others to buy them.” Unsurprisingly, however, they expressed a preference to ride themselves in autonomous vehicles that protect their passengers “at all costs.”

Engineers tend to scoff at the trolley exercise. (“I think the vehicle’s going to stop,” Larry Head told me.) But Shariff says this response takes the problem too literally. “That particular scenario may never come up, but the fundamental ethical trade-off that it’s revealing, apportioning risks among people on the road, is going to happen all the time,” he said. “It already does.” Inching closer to a cyclist in the right lane to move farther from a semi on the left, for instance, prioritizes your safety over the cyclist’s. Leaving that “decision” up to your subconscious becomes its own ethical choice, he added, once “the luxury of deliberation” embodied by the preprogrammed car exists. Human error contributes to an estimated 95 percent of crashes; in theory, automated and connected cars can eliminate our mistakes. “If you recognize that one option is safer,” Shariff says, “and you choose the other option, that’s a moral decision.”

Human error contributes to an estimated 95 percent of crashes; in theory, automated and connected cars can eliminate our mistakes.

At the moment, V2V merely transfers information without dictating

how it is used. But as its reach expands, potentially linking drivers, carmakers and telecommunications companies with state, city and local governments and other private and public interests in a vast network orchestrating the use of what are essentially lethal weapons, questions of responsibility become more complicated. V2V-enabled vehicles would broadcast their size, speed, acceleration and braking activity and location (down to the lane), according to the latest standards. Who will be allowed to access that data? Will it remain anonymous, as automakers currently intend? Who will defend the network against hackers and keep up-to-date the critical software that collects, analyzes and translates that data into best practices for each of its users? Do drivers have the ultimate right to control their cars — even if they want to speed, say — or should the network be programmed to prevent them from doing so? Should it matter whether they’re flooring it because they’re drunk or delivering a pregnant woman in labor to the hospital?

Of course, governments and standards organizations already confront similar utilitarian questions. Early airbags were designed to protect men of average height and weight at the expense of those outside the bell curve; when ambulances run red lights to save a dying patient, they increase the risk of death for everyone around them; a turn-only lane might be safer for bicyclists but worsen vehicle congestion or put a store out of business. Still, each of these decisions can be modified on the micro level by human behavior. The ambulance driver’s experience and his reactions to traffic conditions might mitigate his accident risk. An algorithm applies norms uniformly. Those presets would need to function nationwide, which would require countless entities with conflicting interests and values to submit to a single interpretation of what is best for most people. (Until a majority of cars are fully automated and connected, such a network would most likely behave as the Waze navigation app does, delivering information to human or automated drivers for them to interpret.)

Unfortunately, as a species, humans are notoriously inept at acting in our own long-term interest. Imagine an algorithm for hurricane evacuation that moves cars faster than human drivers would — “balancing the risk of a more-damaging accident from the higher speed against the benefit of getting people out of the evacuation zone faster,” William Whyte, a prominent automotive cybersecurity consultant, wrote to me in an email. “That would be an algorithm that on average worked better than current practice but, when it failed, would fail worse. I think that people would probably feel uncomfortable with that in the same way that they feel uncomfortable on planes — there’s a combination of lack of control plus the fact that if things go wrong, they go very wrong, which triggers all the ways that people are very bad at risk assessment.”

Automation thrives in the absence of regulation; V2V depends on it. The United States Department of Transportation has long helped sponsor V2V research projects: Among the largest is a collaboration between the University of Michigan and the city of Ann Arbor that plans to have 3,150 V2V-equipped cars and 70 connected traffic lights operating there by January. Tampa, New York and the state of Wyoming have won more than $40 million from the agency to implement V2V and other connected-vehicle projects within the next year. At least 41 states have considered legislation surrounding the deployment of automated or V2V technology; at least nine have changed their following-distance laws to permit platooning.

Without federal help, however, the upfront cost of connected infrastructure can be prohibitive for small towns and cities. Bryant Walker Smith, a law professor at the University of South Carolina and an affiliate scholar at Stanford’s Center for Internet and Society, recently challenged a group of students to come up with ways to secure public funding for vehicle-to-infrastructure technology that would enable more governments to afford it. One idea was to use the technology at intersections to collect traffic data and sell it to the developers of automated systems to improve their algorithms; another was to use the dedicated bandwidth in cars to create a meshlike internet network for rural areas, eliminating the need to put down expensive fiber cable.

Perhaps the most provocative idea, Smith told me, was a “crash bounty.” He explained: “We’ve posited that V2I can improve safety — what if you could come up with a scheme for monetizing that? If you could prove you had prevented a crash, you could bill the people who otherwise would have been injured or killed: ‘Good news, we’re 80 percent sure that we saved you from an injury that would have cost you tens of thousands of dollars,

so, insurance company, pay up $40,000 and be happy with that.’ ”

To get a sense of how V2V stands to improve the system as a whole, it helps to understand how cities cope without it. To that end, on a recent Monday morning, just after rush hour, I joined Wilson Aleman and Al Vilar, who have been engineers at the Boston Transportation Department for a cumulative five decades, in the Traffic Management Center on the seventh floor of City Hall. Aleman wore a Red Sox cap. Vilar wore glasses with small eyepieces and thick lenses that gave him the look of a jeweler examining a precious stone.

In front of them, a bank of screens displayed live video of various intersections: cars double-parking on narrow, curving roads; pedestrians disregarding stoplights. Aleman could remember when the video screens used to be a board with a real-time LED map on it — and when the department received postmarked letters of complaint instead of tweets, emails and phone calls. Over the last two decades, Boston has connected 68 percent of its lighted intersections to the Traffic Management Center, allowing engineers to adjust the signaling patterns in real time. The department estimates this has reduced travel times by about 15 percent and emissions and crashes by 8 percent each; it also yields economic benefits of approximately $14 million each year, compared with the $318,000 it cost to implement. Still, much of the traffic data — which helps to determine optimal signal timing as well as the city’s future investments — comes from people logging car movements by hand. Widespread V2V would give traffic departments like Boston’s a wealth of detailed information about key locations. (If all cars had V2V, they could blaze through complex intersections without stopping at all, modifying their speeds and headings to slide into gaps.) Maintenance, however, would be a significant challenge.

Unfortunately, as a species, humans are notoriously inept at acting in our own long-term interest.

Aleman showed me his computer monitor, where the connected lights were represented on a city map as red or green dots. The red dots were ones that weather, construction or local hardware issues had taken offline. He pointed to two reds very close together on Commonwealth Avenue that could strand cars between them and block cross-traffic. “Those are the ones I like to attack,” he said. “Those are important to me.” The phone rang. It was a communications specialist who was elbow-deep in a control box on one of Aleman’s corners, working on a surge protector. “Maybe if you unplug one at a time?” Aleman suggested.

On another phone, Vilar was patiently discussing the timing of a new light on Endicott Street at Keany Square with one of the street’s residents. “We’ll get it so it’s working better; right now it’s working safer because there’s less cars coming at you.” He paused. “They’re ready to march on City Hall, I know.” One-way, residential Endicott begins as an offshoot of I-93 North and ends where the causeway and the Charlestown Bridge converge as six lanes. The light makes it slower to turn left. But, Vilar explained once he’d hung up, “we can’t make it beautiful, because then it will siphon off cars from the causeway.”

He pointed to the Endicott video feed where, sure enough, a line of cars was forming. “See, now it’s like 14 cars deep, people honking — you want it to be more serene, especially when you’re eating dinner,” he said. But say those cars weren’t there, and I-93 was backed up, and Google Maps began suggesting Endicott as a shortcut. “So now you’ve got a guy who just got off there, maybe he’s enraged, instead of going 30 he’s going 40, a kid runs out into the street because he’s not used to the traffic being there. ... ” Vilar said. “You want to make it so it’s moving. You want him to think: I’ll be O.K. I don’t want to get off this street. I’ll stay.” A V2V network could make these sorts of intricate adjustments instantly and indefatigably, but it

would do so without Vilar’s empathy for the harried commuter or the family conversing over supper.

“Sometimes people call,” he said, “and they just want to rant, and you’re just a punching bag for those people, and that’s part of the job.” Vilar grew up in the North End; he, his wife and his children still live nearby. “The woman who just called, she’s almost like an aunt to me,” he said. “She has my number.”

Traffic may be impersonal, but we experience it intimately. Its smells, sounds and movements are as fundamental to our daily passages as the weather. The transition from horses to cars transformed the nature of cities, Ann Norton Greene writes in her 2008 history, “Horses at Work”: “In the name of safety and efficiency, urban Progressives moved children into supervised urban playgrounds, installed new traffic regulation devices, placed policemen on the streets and encouraged changes that turned streets from spaces in a neighborhood to spaces through a neighborhood.” Vehicles, in this milieu, served as a “protected, private space, an extension of the private home.”

More than a century later, in the context of platooning, privacy seems to be able to foster community as much as disrupt it. The dedicated channel over which one truck controls others allows the drivers to converse exclusively with one another, as opposed to using their C.B. radio, which is rife with vulgar chatter. Dave Mercer used to keep his turned off unless he needed news about road conditions. But during the test run, he and Jake Gregory exchanged frequent notes on moves they planned to make or vehicles that the other couldn’t see; Mercer described it as “Jake and I looking out for each other on the road.”

Might the opportunity to collaborate with colleagues in this way reform the macho trucking culture, increasing the diversity of drivers, only 5.1 percent of whom are women? Or might including more varied perspectives in the design process result in even more innovative applications of V2V? The homogeneity at leading technology companies is well documented. Still, it seems notable, considering how infrastructure and power reinforce each other, that of the several dozen engineers, academics and standards developers I spoke with whose thinking is shaping the future of V2V, most were white and all but one was male.

Artificial systems evolve — just as natural systems do — by applying strict rules to events that are often swayed by chance, and the results are frequently surprising. Relinquishing control of our vehicles to algorithms may ultimately transform us from drivers to passengers who tranquilly observe our surroundings from within our bubble or ignore them altogether in favor of watching a movie or checking email. Then again, by the time that sort of algorithm is ready, more and more of us may choose to depend on ride-sharing applications like Uber and Lyft for personal travel and delivery services, or even 3-D printing for goods we once needed a car to go get and pay for. “I have three daughters; they’re totally uninterested in driving,” Jack Pokrzywa, director of global ground vehicle standards for SAE International, told me. “When I got my license, I was 17. I was in seventh heaven.” For those of us for whom a visit to the D.M.V. was a rite of passage, it’s easy to forget that, before Americans found freedom inside a motor vehicle, they found it on foot and astride horses, in covered wagons and trains, on ships, buses, planes, bicycles, roller skates. They found it in pure motion, on a journey to someplace beyond what they could imagine. Where it began and where it ended was still theirs to decide. Likewise, automated connected vehicles stand to make travel safer, maybe even boring, but somebody still has to get in and tell them where to go.